A free, open-source AI chat web app you can run locally in minutes — with your own API key.
Documentation · Community docs · startone.ai
Streaming multi-provider chat plus a watermarked image-generation demo, in one small, readable Next.js codebase. No account, no signup, no database. Bring your own key and go.
StartOne Community is the free tier. It's a taste of the architecture — not a business. When you're ready to build and ship a real AI product, see Upgrading below.
- Streaming AI chat — six models across OpenAI, Anthropic, Google, GLM (z.ai), DeepSeek, and Qwen, via the Vercel AI SDK. Bring your own key with an env var.
- Image-generation demo — a watermarked, 512×512 demo powered by fal.ai. Shows the capability; not for production.
- Runs locally in minutes —
npm install, add a key,npm run dev. No database, no Docker. - Small and readable — one Next.js codebase (React, TypeScript, Tailwind) you can read end to end. No black boxes.
Requires Node 22+.
git clone https://github.com/startoneai/startone-community.git
cd startone-community
npm install
cp .env.example .env.local # add at least one provider key
npm run devOpen http://localhost:3000. With no keys set, the app still boots: chat returns labeled dry-run replies and the image demo shows a set-your-key notice. Add a key to turn a feature on.
Set keys in .env.local (all optional — add the ones you want):
| Variable | Enables |
|---|---|
OPENAI_API_KEY |
Chat with GPT models |
ANTHROPIC_API_KEY |
Chat with Claude models |
GOOGLE_GENERATIVE_AI_API_KEY |
Chat with Gemini models |
ZAI_API_KEY |
Chat with GLM models |
DEEPSEEK_API_KEY |
Chat with DeepSeek models |
DASHSCOPE_API_KEY |
Chat with Qwen models |
FAL_KEY |
The image-generation demo (fal.ai) |
Keys stay local — they're read from the environment only. There's no database, and nothing is sent anywhere except directly to the provider you configured.
Full configuration reference: docs.startone.ai/community.html
StartOne Community is deliberately simple: single-user, local, bring-your-own-key. Chat history is session-only. The image demo is watermarked and capped at 512×512 on purpose. There's no billing, no auth, no persistence.
It's the free taste of a larger system — a clean, honest starting point to explore the architecture.
Ready for more than a demo? The paid editions are the same architecture, grown up:
- StartOne Developer — Web ($249) — the full boilerplate: media studio (image, video, audio), gallery, multi-model catalogue, Clerk auth, Postgres, Redis, storage, and a deploy-ready setup.
- StartOne Developer — Message ($249) — one AI bot across eight messaging platforms: Discord, Telegram, Slack, WhatsApp, WeCom, Feishu, DingTalk, and WeChat OA.
- StartOne Business — launch and charge customers: billing, a prepaid credit ledger, payment rails, and multi-tenancy.
The two Developer editions are sold separately, because most developers build either a web app or a bot — not both.
Full documentation for every edition — setup, every environment variable, the architecture, and deployment — is at docs.startone.ai. Read the manual before you buy.
Learn more and upgrade at startone.ai.
MIT — see LICENSE. Free to use, modify, and build on.
StartOne™ is a trademark of Kredit Technology, LLC. The MIT license covers the source code; it does not grant rights to use the StartOne name or logo to brand your own product or to imply endorsement.
Community support via GitHub Issues and Discussions. For the paid editions, support is included with your purchase at startone.ai.